These are questions I suggested people ask themselves after reading my post on Anthropic.
When I was writing the post back in November, I did not know about their RSP plans. I did, though, try to communicate my understanding of the leadership’s thinking: the moment a commitment prevents Anthropic from training or deploying a model, they would get rid of that commitment.
Fundamentally, decision-making at Anthropic has never been based on such commitments; these commitment are a PR instrument, not something that would actually bind what the structure of Anthropic is.
I am sad this happened; I would be much happier in a world where I was proven, again and again, wrong about all of my AI-related beliefs; but also, I hope people at Anthropic are capable of taking the outside view and updating.
I wonder if it perhaps could be helpful to imagine yourself at OpenAI, believing Sam Altman, who you know to be very good at telling you what you want to hear, and trying to figure out where to find evidence for what’s actually true about OpenAI’s decision-making and governance.
These are questions I suggested people ask themselves after reading my post on Anthropic.
When I was writing the post back in November, I did not know about their RSP plans. I did, though, try to communicate my understanding of the leadership’s thinking: the moment a commitment prevents Anthropic from training or deploying a model, they would get rid of that commitment.
Fundamentally, decision-making at Anthropic has never been based on such commitments; these commitment are a PR instrument, not something that would actually bind what the structure of Anthropic is.
I am sad this happened; I would be much happier in a world where I was proven, again and again, wrong about all of my AI-related beliefs; but also, I hope people at Anthropic are capable of taking the outside view and updating.
I wonder if it perhaps could be helpful to imagine yourself at OpenAI, believing Sam Altman, who you know to be very good at telling you what you want to hear, and trying to figure out where to find evidence for what’s actually true about OpenAI’s decision-making and governance.
Apply that to Dario and to Anthropic.